Confidence interval estimation using linear combinations of overlapping variance estimators

We develop new confidence-interval estimators for the mean and variance parameter of a steady-state simulation output process. These confidence intervals are based on optimal linear combinations of overlapping estimators for the variance parameter. We present analytical and simulation-based results exemplifying the potential of this technique for improvements in accuracy for confidence intervals.

• Type: Conference
• Year: 2007
• Where: WSC
• Authors: Tûba Aktaran-Kalayci, David Goldsman, James R. Wilson
• Keywords: Confidence Intervals, Modeling And Simulation, Steady-state Simulation Output, Variance Parameter, WSC 2007